Morphological variation of an ornament expressed inboth sexes of the mosquito Sabethes cyaneus

Sandra H. South and Göran Arnqvist

Department of Animal Ecology, Evolutionary Biology Centre, Uppsala University, Uppsala, Sweden

ABSTRACT

Question: Do elaborate ornaments expressed in both sexes show patterns of phenotypicvariation consistent with selection via mutual mate choice?

Data description: Detailed morphometric data on the striking leg ‘paddle’ ornament ofthe sabethine mosquito Sabethes cyaneus: ornament size and shape and size of generalmorphological traits. Data derive from 80 males and 80 females from a wild-type laboratorycolony established with individuals collected in Panama.

Search method: Shape variation was analysed using geometric morphometric methods(elliptic Fourier analyses). We investigated sex differences in the relationships between body sizeon the one hand and leg length, ornament size, and ornament shape on the other, using generallinear models. We also explored morphological variation in asymmetry, allometry, and in themagnitude of phenotypic variation across traits.

Conclusions: These ornaments showed many of the classic hallmarks of a sexually selectedand condition-dependent ornament: (i) phenotypic variation in size was much greater than forany other trait; (ii) the size of the major part of the paddle showed positive allometry; and(iii) the degree of fluctuating asymmetry in one component of the shape of the leg paddlesdecreased with body size. Remarkably, these patterns were more pronounced in females andsexual dimorphism in size and shape of the leg paddle ornament was slight. Although data onthe current pattern of morphological variation alone does not allow firm conclusions aboutpast selection, our results are consistent with the maintenance of these ornaments in both sexesby sexual selection via mutual mate choice.

Keywords: allometry, Diptera, elliptic Fourier analysis, sexual selection, signal, variation.

INTRODUCTION

The strikingly elaborate ornaments (defined as exaggerated or novel structures used tovisually attract mates) possessed by many animal species have inspired much research intothe evolution of signals via sexual selection. Most of these studies have been conductedin avian taxa with a particular focus on the elongated, curled, and often brightly colouredtail feathers (Darwin, 1871; Jennions, 1993; Andersson, 1994; Cuervo and Møller, 1999). Factors that affect the

Correspondence: S.H. South, Department of Animal Ecology, Evolutionary Biology Centre, Uppsala University,Norbyvägen 18D, SE-752 36 Uppsala, Sweden. e-mail: sandra.south@ebc.uu.seConsult the copyright statement on the inside front cover for non-commercial copying policies.

Evolutionary Ecology Research, 2009, 11: 1–21

© 2009 Sandra H. South

intensity of sexual selection include gamete investment, mating investment, parentalinvestment, and variance in mate quality (see Andersson, 1994; Shuster and Wade, 2003). In the majorityof avian species, females invest a relatively higher amount into gamete production (Trivers,

1972). Moreover, both females and males are known to mate multiply and seek extra-paircopulations, further increasing the variance in male fitness and intensifying the strengthof sexual selection (Ligon, 1999). This characterizes the common scenario of relativelyindiscriminate males competing for mating opportunities with choosy females, which formsthe general basis for our understanding of the evolution of ornaments in males by femalemate choice (Ligon, 1999; Dunn et al., 2001).

Our understanding of ornament evolution in insects is much more restricted. There aremany descriptions of exaggerated male structures that represent putative ornaments, as theyare displayed during courtship (see Thornhill and Alcock, 1983; Sivinski, 1997). However, it generallyremains unknown which factors have contributed to the evolution of these ornaments.One notable exception is the evolution of antlers and eye stalks in flies. Although femalepreference for these ornaments has been demonstrated (Burkhardt and de la Motte, 1988; Wilkinson and

Reillo, 1994; Hingle et al., 2001; Cotton et al., 2006), they are employed in male–male competition andintrasexual selection clearly plays a key role in their evolution (Wilkinson and Dodson, 1997). Ingeneral, scramble competition polygyny is considered to be widespread in many insect taxa(Thornhill and Alcock, 1983). This has contributed to the widely held view that there is littleopportunity for female mate assessment and direct female choice in insects (Alexander et al.,

1997), although indirect female choice (sensu Wiley and Poston, 1996) is common. Yet, there isgrowing evidence that female insects indeed assess males and show direct mate choice [e.g.burying beetles (Beeler et al., 2002), stalk-eyed flies (Burkhardt and de la Motte, 1988), wax moths (Jang and

Greenfield, 1996, 1998), and field crickets (Simmons, 1986)]. In most insects, both males and femalesmate multiply and exhibit no parental care (Thornhill and Alcock, 1983; Arnqvist and Nilsson, 2000).Here, the selective pressures acting upon ornaments may be largely understood in terms ofdifferential investment into gamete production. However, in other insects the mating systemand pattern of ornamentation is more atypical, and these systems can provide valuableinsights into ornament evolution by sexual selection.

Our study organism, the sabethine mosquito Sabethes cyaneus (Diptera: Culicidae),possesses elaborate paddle-like ornaments consisting of elongated and iridescent scales onthe mid-leg tarsus and tibia (Fig. 1). Four facts point to a central role for sexual selectionin the evolution of these visually striking paddles. First, unlike most mosquitoes, sabethine

Fig. 1. The leg paddle of Sabethes cyaneus. The tibial (right) and tarsal (left) component of thisornament are separated by a small indentation.

South and Arnqvist2

mosquitoes are diurnal and show unique visual adaptations in the optical structure of theireyes (Land et al., 1999). The fact that their visual ornaments are also unique among culicidmosquitoes (Sivinski, 1997) is therefore consistent with a signal function. Second, there is largevariation in paddle number, shape, and coloration, but not in general morphological traits,across closely related sympatric sabethine species (Harbach and Petersen, 1992; Judd, 1996). Such apattern of rapid morphological divergence is a general hallmark of sexual selection (Coyne and

Orr, 2004). Third, complex male courtship behaviours, during which these mid-leg ornamentsare prominently displayed in front of the female, have been described in all four species ofsabethines studied to date, including our study system, S. cyaneus (Okazawa et al., 1986; Hancock et

al., 1990a; Philips et al., 1996; Zsemlye et al., 2005). Fourth, direct phenotypic engineering experiments(i.e. removal of the ornaments) have shown that this ornament affects mating successin S. cyaneus, yet flight and oviposition behaviours remain unaffected (Hancock et al., 1990b).Notably, this effect was stronger, and only statistically significant, in females.

Sabethes cyaneus provides a particularly intriguing model system. This is not onlybecause the ornaments are extraordinarily striking (Shannon, 1931), but also because theornaments appear to be sexually monomorphic to the naked human eye and becausefemales are strictly monandrous while males are polygynous (South and Arnqvist, 2008; S.H. South

et al., submitted). In addition, males and females adopt traditional sex roles, with males searchingfor females perched on horizontal branches before aligning to perform a highly stereotypedand vigorous courtship dance in which the ornaments are displayed in front of the female(Hancock et al., 1990a). Furthermore, females are generally reluctant to mate, there is no parentalcare by either sex, and there is no evidence of nutritional gifts being transferred to femaleswith the male seminal fluid. Given the mating system of S. cyaneus, the occurrence of maleornaments (henceforth referred to as ‘leg paddles’) is entirely consistent with evolutionarytheory. However, it is unclear why we should see the expression of ornamental leg paddles infemales.

There are two main classes of hypotheses that aim to explain the expression of elaborateornaments in both sexes (for a review, see Kraaijeveld et al., 2007). The first posits that male and femaleexpression of the ornament is due to direct selection on the ornament in both sexes. Directselection may result from (i) selection for sexual ambiguity in females (Burley, 1981; Langmore and

Bennett, 1999), (ii) female contest competition (West-Eberhard, 1983; Gwynne, 1991), and/or (iii) mutualmate choice (Huxley, 1914; Kraaijeveld et al., 2004). The second group of hypotheses suggests thatthe expression of the trait in females is a non-adaptive result of an intersexual geneticcorrelation, such that the female ornament is merely a correlated response to selection forthe ornament in males (Darwin, 1871; Lande, 1980, 1987; Lande and Arnold, 1985). By simply consideringwhat is already known of the morphology and mating system of our model system, severalof these hypotheses can be rejected.

Selection for sexual ambiguity could occur when being male brings social benefits (Burley,

1981) or reduces sexual harassment of females (Cook et al., 1994; Forbes et al., 1997; Van Gossum et al., 2007).It is highly unlikely that there are any social benefits such as access to food sites that couldselect for sexual ambiguity in S. cyaneus, simply because mosquitoes do not exhibit socialbehaviour. A reduction in the amount of sexual harassment is also unlikely, as males ofS. cyaneus court males as eagerly as females (Hancock et al., 1990a; personal observation), as do malesof many insect species,. More importantly, experimental removal of the leg paddles infemales actually led to a reduction in courtship by males (Hancock et al., 1990b). These factsstrongly suggest that the maintenance of these ornaments in females cannot be accountedfor by selection for sexual ambiguity. Furthermore, no aspect of the ecology of this species

Allometry of an ornament expressed in both sexes 3

could generate female contest competition. Sites for oviposition (plant-held waters),mating (horizontal branches), and feeding (primate hosts, flowers) are abundant in naturalpopulations, females do not actively seek males, and females have never been observed toengage in contests. Thus, female contest competition is very unlikely to contribute to theelaboration of these ornaments in females.

The remaining applicable hypotheses that may account for the maintenance of theseornaments in females are direct selection via mutual mate choice or non-adaptive expres-sion due to an intersexual genetic correlation. Theory has shown that mutual mate choicecan evolve under a range of conditions and studies of monogamous birds support thepresence of selection by mutual mate choice (Jones and Hunter, 1993, 1999; Kraaijeveld et al., 2004; Hooper

and Miller, 2008). However, S. cyaneus adopts ‘traditional’ sex roles: male polygyny/femalemonandry, in combination with male mate searching and male courtship/female materejection. Nonetheless, there is growing empirical support for male/mutual mate choice alsoin polygynous systems (for reviews, see Amundsen, 2000; Bonduriansky, 2001; Kraaijeveld et al., 2007), which hassparked recent theoretical interest in this area (Servedio and Lande, 2006; Servedio, 2007). It is alsopossible that the occurrence of female paddles in S. cyaneus is due solely to a correlatedexpression in females of sexual selection for paddles in males. To our knowledge, there areno known cases of non-adaptive expression of elaborate ornaments to the extent that theyare basically sexually monomorphic. This fact has been accredited to the costs associatedwith ornament expression in females, such as increased conspicuousness to predators (Godin

and McDonough, 2003), impaired ability and energetic costs during flight (Evans, 2004; Allen and Levinton,

2007), and production costs during ontogeny (Munoz et al., 2008). However, it remains possiblethat the costs associated with expression of these paddles in S. cyaneus females are lowenough to permit such an extreme and intriguing case of non-adaptive expression. We note,however, that female paddles do affect female attractiveness to and/or recognition by males,as their removal results in a drastic reduction in female mating rate and courtship attemptsby males (Hancock et al., 1990b).

One potentially useful route to gain insights into the evolution of ornaments in any taxais to study the pattern of ornament expression. Here, studies comparing the variability ofornaments and other body parts are valuable (e.g. Schluter and Price, 1993), as are studies of staticallometry of ornaments (e.g. Kodric-Brown et al., 2006). This approach is particularly suitable foradult holometabolous insects, because they grow solely during egg, larval, and pupal stages(Klingenberg and Zimmermann, 1992; Emlen, 1994; Kemp and Rutowski, 2007). In this study, we explore thepattern of phenotypic variation of the leg paddle ornament in S. cyaneus by assessingvariability as well as the relationships between ornaments and size in both sexes. We employgeometric morphometric methods, which allow for powerful analyses of variation in sizeand shape (Klingenberg, 1996).

One of our goals is to assess those facets of morphological variation and allometry thatmay help discriminate between direct selection via mutual mate choice and an intersexualgenetic correlation. We begin by exploring sexual dimorphism in general body measures aswell as ornament shape and size. Under mutual mate choice, we expect directional selection(see below) in both sexes to lead to similar expression in males and females, even ifsex-specific factors are likely to affect expression to some extent (Kimball and Ligon, 1999). Asdiscussed above, however, a non-adaptive genetic correlation may also lead to similarexpression in males and females, provided that there are no costs of expressing theornament in females. We then proceed to explore the following three classic hallmarks ofsexual selection. First, sexual ornaments are expected to be more phenotypically variable

South and Arnqvist4

than non-sexual traits (Schluter and Price, 1993; Rowe and Houle, 1996; Cuervo and Møller, 1999; Badyaev,

2004). Second, the degree of bilateral asymmetry in sexual ornaments may decrease withbody size, as a result of condition-dependent expression (e.g. Swaddle, 2003; but see Cotton et al., 2004).Third, sexual ornaments often, but not always (Bonduriansky, 2007), show positive staticallometry (i.e. increase with overall body size at a rate > 1.0) presumably as a reflection ofdifferential trade-offs in resource allocation during ontogeny (Grafen, 1990a, 1990b; Green, 1992;

Petrie, 1992; Emlen and Nijhout, 2000; Knell et al., 2004; Kodric-Brown et al., 2006).Although several models have suggested that sexually selected traits should show positive

allometry (see references above), the generality of this prediction was critically appraised byBonduriansky and Day (2003). They showed that the classic sexual selection scenario doesnot necessarily result in positive allometry. A recent reappraisal by Bonduriansky (2007)

delineated the conditions that should lead to positive allometry and we suggest that thoseconditions are met in S. cyaneus. First, the ornament is truly a ‘dedicated’ secondary sexualcharacter that does not have any known viability-related function (see above). Second,there is no known alternative male mating tactic that may favour small ornaments andresult in disruptive selection. Third, paddle expression does not appear to affect flight abilityor oviposition behaviour (see above; Hancock et al., 1990b) and it thus is unlikely that intermediatetrait sizes are favoured for such reasons. Fourth, it is very reasonable to assume thatthe relative viability costs of expressing this ornament diminishes with body size (seeDiscussion).

MATERIALS AND METHODS

We used a strain of S. cyaneus established by R.G. Hancock and W.A. Foster in December1988 from a sample of mated females collected at the Isla de Maje, Lago Bayano, Panama,Republic of Panama. This colony was maintained at Ohio State University, USA at apopulation size of approximately 200–300 individuals. Our colony has been housedat Uppsala University, Sweden since April 2006 at 26 ± 1�C, 78–82% relative humidity, anda 12L:12D photoperiod, at a population size of approximately 400 individuals. Larvaewere reared in plastic trays (21.5 × 14.5 × 5 cm) filled to 2.5 cm with deionized water, whichwas changed weekly. They were fed an ad libitum diet of crushed fish flake food. Pupae werecollected in small dishes (diameter 8 cm; height 2.5 cm) and these were placed in terraria(29 × 17.5 × 18 cm). An ad libitum supply of honey-soaked sponges and deionized waterwicks was provided in these terraria.

We collected and froze adults within 24 h of emergence to avoid variation in leg paddlesthat is due to wear and/or damage. Individuals were sampled across three generationsall raised under the rearing regime described above. We then collected morphometricinformation using a digitizing tablet (Summasketch® III) placed under a side-mountedcamera lucida attached to a dissecting microscope (Leica® MZ8). The following traitswere measured in a sample of 80 females and 80 males on the left side of the body: winglength; thorax width; proboscis length; antenna length; fore-leg femur, tibia, and tarsuslength; mid-leg femur, tibia, and tarsus length; hind-leg femur, tibia, and tarsus length.Wing length was used as an integrative measure of body size (Siegel et al., 1992; Armbruster and

Hutchinson, 2002). We also captured the outline of both the left and right leg paddles in twoseparate segments: tibial paddle and tarsal paddle. All measurements were taken twice toenable an estimation of the repeatability of our measures in a subset of 30 males and30 females.

Allometry of an ornament expressed in both sexes 5

All length measurements were converted to millimetres before analysis. The leg paddleoutlines were analysed using elliptic Fourier analysis (EFA) (Ferson et al., 1985; Rohlf, 1992). Theoutlines of all paddles for all individuals were included in a common EFA, using thesoftware Morpheus et al. (Slice, 2002), which was also used to extract the area of all paddles.The EFA itself was made invariant of size, position, and rotation, and used 12 harmonics(yielding 48 Fourier coefficients). These functions provided a near perfect fit to all paddleoutlines. To reduce the dimensionality of our shape descriptors, the 48 Fourier coefficientsfor each leg paddle segment were treated as variables in principal component analysesperformed on the covariance matrix (Rohlf and Archie, 1984). The first five principal components(PCs) from each leg paddle were retained for subsequent analyses. These five PCscollectively described 88.79% and 87.53% of the shape variation in tibial and tarsal paddlesrespectively (see Appendix 1).

The relationship between paddle area and length measurements within and between thesexes was investigated using full analyses of covariance (ANCOVAs) with sex as a fixedfactor and length measurements as a continuous covariate. Analogously, paddle shape wasinvestigated using multivariate analyses of covariance (MANCOVAs) with the five PCs asdependent variables, sex as a fixed effect factor, and body size as a covariate.

We compared the magnitude of phenotypic variation in traits presumed to be undersexual selection (leg paddle area and the length of the leg segments the paddles are locatedon – mid-leg tarsus and tibia) and other representative body traits (thorax, wing, fore-legfemur, tibia, and tarsus) in two different ways. First, we compared the coefficients ofvariation across traits. Second, following a transformation where all values were divided bythe trait mean, we compared the proportional variances across traits. We note that the lattermethod gives all traits a mean value of one and a variance that is proportional across traits(Sokal and Rohlf, 1995).

We calculated leg paddle asymmetry simply by subtracting the left-hand paddlevalue from the right-hand paddle value for each individual and for all measures of paddlemorphology, in both tibial and tarsal paddles. We assessed whether the pattern ofasymmetry for all leg paddle variables was consistent with directional asymmetry[as opposed to fluctuating asymmetry (Swaddle, 2003)] by conducting t-tests of H0: µ = 0 (whereµ is the mean signed asymmetry), both for the entire sample and for males and femalesseparately. The relationship between body size and unsigned asymmetry was then assessedby Spearman rank correlations between the absolute value of asymmetry and bodysize. Any unsigned asymmetry variables with a significant correlation were then furtherinvestigated using ANCOVAs with sex as a fixed factor and body size as a covariate.

The allometric slopes were estimated using Reduced Major Axis regression [RMA,Model II regression or Geometric Mean regression (Ricker, 1984; McKinney and McNamara, 1991)] ofbody size against the paddle size and leg measurements, using the software MODEL II(Sawada, 1999). For paddles, the area measurements were first linearized using an ellipsetransformation as the paddle resembles an ellipse. We used the equation describing the areaof an ellipse (A = π × a × b, where a and b are the lengths of the ellipse’s semi-major andsemi-minor axes). We approximated a to be 2 × b and used this to solve for the length of themajor axis. We henceforth denote this linearized measure of paddle area as ‘paddle size’.This linearized value of paddle area was then converted to millimetres to make it directlycomparable to other length measurements. Before the RMA regressions, all data were log(1 + x) transformed.

South and Arnqvist6

RESULTS

Most of our measurements were highly repeatable (r > 0.70; Appendix 2). It is noteworthythat this was also true for our multivariate measures of paddle shape and for measures ofasymmetry. Thorax width, proboscis length, and antennal length measurements showed aslightly lower repeatability than wing length and will not be used in the subsequent analyses.We note that females were larger than males (t 144 = 9.353, P < 0.001, based on separatevariances).

We investigated the relationship between tibial paddle size and body size within andbetween the sexes. Tibial paddle size increased with body size in both sexes (F1,156 = 35.3,P < 0.001) (Fig. 2a), but there was no main effect of sex (F1,156 = 0.6, P = 0.442) and nosignificant interaction between tibial paddle size and body size (F1,156 = 0.7, P = 0.421). Incontrast, the relationship between body size and tarsal paddle size differed significantlybetween the sexes (Fig. 2b). As for tibial paddle size, tarsal paddle size increased with bodysize (F1,156 = 16.7, P < 0.001), but there was also a main effect of sex (F1,156 = 4.4, P = 0.039)and, most importantly, an interaction between tarsal paddle size and body size (F1,156 = 5.6,P = 0.019). This relationship was steeper in females than males, revealing that paddles wereproportionately larger in females (Fig. 2b).

The relationship between body size and mid-leg tibia and tarsus length was similar to thatobserved with the part of the paddles situated on each corresponding segment. That is, therelationship between body size and mid-leg tibia did not differ between the sexes, whereasthat between body size and mid-leg tarsus did (Table 1), with females having proportionatelylonger tarsi. The fore-leg tibia and tarsus length showed a similar pattern to the mid-legtibia and tarsus: female fore-leg segments were proportionately longer than those of males(Table 1). Notably, the hind-leg tibia and tarsus length differed from this pattern ofproportionately larger female paddles: although the relationship differed between the sexes(Table 1), both male hind-leg tibia and tarsus length were proportionately longer than forfemales.

The relationship between leg paddle shape and leg paddle size differed between the sexes(see Table 2) and, in this sense, this ornament is sexually dimorphic. These effects werelarger for the tarsal components of the leg paddles, and a closer analysis using univariateF-tests of each of the tarsal paddle shape PCs showed that the main difference between thesexes falls on the first principal component, PC1 (sex: F1,156 = 8.4, P = 0.004; tarsal paddlearea: F1,156 = 242.2, P < 0.001; sex × tarsal paddle area: F1,156 = 32.6, P < 0.001). Visualiza-tions of this result showed that paddles change shape more rapidly with paddle size in malesthan in females (Fig. 3a). For the first PC of tarsal paddle shape, male paddles tend to beslightly broader at the widest point than do female paddles and the breadth of the paddleincreases with paddle area (Fig. 3b). In summary, although these analyses show that paddleshape differs significantly between the sexes, the differences are very slight in absolutemagnitude.

We tested for equality of variances across all traits (using variables transformed to showproportional variances – see Methods) and found significant differences in variability acrosstraits (Bartlett’s test: χ2 = 1259.3, P < 0.001). Visual examination of the data showed thatpaddle size was more variable than any other trait (Fig. 4). Interestingly, although thevariation in the leg segments that carry the paddles is also relatively high, they do notshow the same strikingly high levels of variation as the ornaments themselves (see Fig. 4).Additionally, for many traits females showed significantly higher phenotypic variation than

Allometry of an ornament expressed in both sexes 7

Fig. 2. The relationship between body size and (a) tibial paddle area and (b) tarsal paddle area. Lines(solid: females; dotted: males) represent ordinary linear regressions.

South and Arnqvist8

Table 1. Analyses of covariance of the relationshipbetween body size and leg segment lengths in the twosexes

F d.f P

Mid tibia lengthSex 3.841 1, 156 0.052Body size 61.953 1, 156 0.000Sex × body size 3.914 1, 156 0.050

Mid tarsus lengthSex 5.466 1, 155 0.020Body size 36.942 1, 155 0.000Sex × body size 9.171 1, 155 0.003

Fore tibia lengthSex 15.923 1, 156 0.000Body size 111.077 1, 156 0.000Sex × body size 14.334 1, 156 0.000

Fore tarsus lengthSex 5.356 1, 156 0.022Body size 62.873 1, 156 0.000Sex × body size 5.307 1, 156 0.023

Hind tibia lengthSex 11.852 1, 156 0.001Body size 112.720 1, 156 0.000Sex × body size 8.862 1, 156 0.003

Hind tarsus lengthSex 9.475 1, 156 0.002Body size 75.339 1, 156 0.000Sex × body size 8.638 1, 156 0.004

Table 2. Multivariate analyses of covariance of paddle shape

Wilks’ λ F d.f P

Tibial paddleSex 0.889 1.842 10, 147 0.058Tibial paddle area 44.519 44.519 10, 147 0.000Sex × tibial paddle area 0.794 3.804 10, 147 0.000

Tarsal paddleSex 0.873 2.144 10, 147 0.024Tarsal paddle area 0.297 34.767 10, 147 0.000Sex × tarsal paddle area 0.774 4.287 10, 147 0.000

Allometry of an ornament expressed in both sexes 9

males. These patterns were very similar whether variability was measured as proportionalvariance or as the coefficient of variation (see Fig. 4). We note (i) that the elevated variationseen in paddle size was clearly not due to inflated measurement error, as the repeatabilities

Fig. 3. (a) The relationship between tarsal paddle area and shape, represented by the first principalcomponent of variation in tarsal paddle shape. Lines represent ordinary linear regressions. (b) Theupper and lower extremes of the first principal component of tarsal paddle shape. For further details,see text.

South and Arnqvist10

for ornament size (range: 0.90–0.98) were actually higher than for most other traits(see Appendix 2), and that (ii) the exact same pattern is seen when analysing phenotypicvariation in paddle area rather than paddle size.

The only measure of asymmetry of leg paddle morphology that showed a significantrelationship with body size was PC2 of tarsal paddle shape [rs = −0.253, P < 0.05, using falsediscovery rate compensation (Storey, 2003)]. An ANCOVA confirmed the negative relationshipbetween the unsigned asymmetry of PC2 and body size (F1,156 = 7.6, P = 0.007), and showedthat the sexes did not differ significantly in this respect (sex: F1,156 = 1.5, P = 0.217;sex × body size: F1,156 = 1.5, P = 0.223). Hence, although no other measure of paddleshape and size showed a significant correlation with body size, these analyses show thatpaddles do grow slightly more symmetrical with size. We also note that the patternof asymmetry in leg paddles was consistent with fluctuating, rather than directional,asymmetry (Appendix 3).

We found positive allometry for tibial paddle size in both males and females (see Table 3).There was, however, no significant difference from isometry for tarsal paddle size. Notably,tarsal leg segment length (carrying the tarsal component of the paddle) did show positiveallometry in both sexes. In contrast, there was little evidence for deviations from isometryfor fore- and hind-leg segments.

DISCUSSION

Studies of the pattern of morphological variation of sexual ornaments have a long history(Huxley, 1932), and many authors have suggested that one can gain insights into development,

Fig. 4. Phenotypic variation of morphological traits, shown both as proportional variances andcoefficients of variation (see text). Traits for which the proportional variance differed significantlybetween the sexes (P < 0.005) are indicated with a star.

Allometry of an ornament expressed in both sexes 11

past selection, and evolutionary history from the extant pattern of trait variation alone(see McKinney and McNamara, 1991). However, it is now clear that the strength of such inferencesis somewhat limited. For example, several models have predicted that ornaments shouldoften show positive allometry (Grafen, 1990a, 1990b; Kodric-Brown et al., 2006). Yet, as outlined in theIntroduction, Bonduriansky and Day (2003) clearly showed that the classic sexual selectionscenario does not necessarily result in positive allometry. Nevertheless, positive allometry isa characteristic of ornamental traits used to visually attract mates in many taxa (see Kodric-

Brown et al., 2006) and a number of important insights have been gained from studies ofphenotypic variation of ornamental traits. In insects, studies of variation and allometryhave been key in understanding the evolution and maintenance of sexually selected‘ornamental’ traits in, for example, stalk-eyed flies (Wilkinson and Dodson, 1997), horned beetles(Emlen et al., 2007), earwigs (Simmons and Tomkins, 1996), and butterflies (Kemp, 2006). Below, we firstask whether the pattern of trait variation in the leg paddles of both sexes of S. cyaneusconforms to the predictions under mutual mate choice. We then discuss how compatible ourresults are with the possibility that ornament expression in females is the result of an

Table 3. Estimates of allometric scaling coefficients of leg paddles and leg segment lengths in thetwo sexes

Trait Sex Slope SE rp t

Tibial paddle size Female 1.931 0.183 0.528 5.080Male 2.164 0.226 0.360 5.155

Tarsal paddle size Female 0.890 0.087 0.494 1.270Male 0.931 0.103 0.136 0.674

Fore femur Female 0.992 0.068 0.791 0.110Male 0.820 0.072 0.624 2.512

Fore tibia Female 1.125 0.076 0.800 1.661Male 0.919 0.092 0.445 0.874

Fore tarsus Female 1.219 0.105 0.637 2.087Male 0.933 0.094 0.443 0.717

Hind femur Female 1.067 0.076 0.768 0.880Male 0.913 0.088 0.511 0.997

Hind tibia Female 1.107 0.077 0.782 1.388Male 0.915 0.088 0.504 0.960

Hind tarsus Female 1.786 0.149 0.666 5.276Male 1.223 0.121 0.470 1.845

Mid femur Female 1.078 0.072 0.801 1.085Male 0.889 0.081 0.579 1.363

Mid tibia Female 1.110 0.094 0.653 1.172Male 0.982 0.099 0.435 0.187

Mid tarsus Female 1.469 0.135 0.573 3.485Male 1.700 0.189 0.088 3.697

Note: Traits that show significant allometry (i.e. β ≠ 1; t-test of H0: β = 1) are indicated in bold. SE = standard error.

South and Arnqvist12

intersexual genetic correlation, and finish with what we consider to be the most importantimplications of our findings.

The ornamental leg paddles of S. cyaneus showed many of the classic hallmarks of acondition-dependent (sensu Rowe and Houle, 1996) ornament under directional sexual selection inboth sexes. First, variation in the size of the leg paddles was much greater than in generaltraits. While the latter showed a degree of phenotypic variation typical for morphologicaltraits (CV: 2–7%), the leg paddles showed a variation (CV: 12–17%) that was even greaterthan the typical ornament in birds (Cuervo and Møller, 1999). Second, the degree of fluctuatingasymmetry of one aspect of the shape of the tarsal paddle was negatively related to size, asmight be expected for condition-dependent ornaments (Watson and Thornhill, 1994; Swaddle, 2003; Polak

and Starmer, 2005). Third, in contrast to most general traits, the major part of the ornamentalleg paddles showed a strongly positive allometric scaling coefficient (∼2).

However, we note that our results are not fully in accordance with the predicted patternsfor a sexually selected ornament. First, the smaller of the two leg paddle components, thetarsal paddle, did not show positive allometry. However, the length of the tarsal segmentupon which this ornament component is situated did show positive allometry. As thissegment is itself an integrated part of the ornament (see Fig. 1), it is reasonable to never-theless conclude that the ornament as an integrated structure showed positive allometry.Second, a trait that we do not consider to be under sexual selection, the length of the hindtarsus in females, also showed positive allometry. It is worth noting that this pattern wasagain seen on the tarsus, albeit of a leg that does not carry the ornament. We suggest thatthis may result from developmental integration of tarsi across legs. Indeed, regulationof tarsal development in Drosophila has been shown to be dependent on a molecularmechanism that is distinct to the other leg segments (Campbell and Tomlinson, 1998; Rauskolb, 2001).Third, as mentioned in the previous paragraph, only the tarsal paddle showed a significantnegative relationship between fluctuating asymmetry and size and this was only true foran aspect of paddle shape that describes a sub-dominant form of paddle shape variation(i.e. PC2). Unfortunately, we did not explore environmental effects on trait expression andthus we are restricted in our ability to draw detailed inferences as to the presence or lack offluctuating asymmetry in particular aspects of ornament morphology (Polak and Starmer, 2005).We note, however, that no aspect of fluctuating asymmetry of the ornament showed apositive relationship with size.

Despite these departures from ideal conformity with predictions, the results are generallyconsistent with the hypothesis that the leg paddles of both males and females are sexuallyselected. This is further supported by the fact that the ornament varied little in size andshape between the sexes and had similar allometric scaling coefficients in males and females.If anything, leg paddles tended to be proportionally larger and more variable in females andleg paddle area increased more rapidly with body size in females than in males. We stressthat the pattern we document here is strikingly different from that of most other taxa, andwe know of no comparable findings in any taxa. In birds with some female expression ofmale ornaments that are assumed to be under sexual selection only in males, females showless variation and a much lower allometric scaling coefficient than do males (see Cuervo and

Møller, 1999). Thus, our results suggest that the expression of this ornament is at least ascondition dependent in females as it is in males. Unfortunately, we know of no otherdetailed morphometric analyses of ornaments expressed in both sexes that are thought tobe used in mutual mate choice.

Two facets of our results are particularly difficult to reconcile with the hypothesis that the

Allometry of an ornament expressed in both sexes 13

expression of leg paddles in S. cyaneus females represents a non-adaptive correlatedresponse to sexual selection in males only. First, the pattern of ornament variation infemales showed the classic hallmarks of directional sexual selection to a greater, rather thanlesser, extent compared with males. This is in stark contrast to bird systems where femaleornament expression is thought to be a non-adaptive correlated response to sexual selectionin males (Cuervo and Møller, 1999). Second, the fact that the leg paddle ornament was basicallysexually monomorphic would only be predicted if trait expression was cost-free in females.However, there should be at least some size-dependent viability costs of this ornament,because both costs of growth (allocation of cuticular material during metamorphosis) andmaintenance [much time is spent grooming (personal observation)] of leg paddles should becondition dependent to some extent. In addition to the present results, the fact that experi-mental removal of the paddles has a larger, rather than smaller, effect on mating rate infemales than males (Hancock et al., 1990b) is clearly not expected if the ornament is only undersexual selection in males. Although these considerations imply that it is very unlikely that anintersexual genetic correlation is responsible for ornament expression in female S. cyaneus,it remains theoretically possible that this is the case. If, for example, the costs of ornamentexpression are much lower in females, sex-specific natural selection could interact with acorrelated response and result in proportionately larger and more variable leg paddlesin females. However, if this was true, our model system would provide the most extremecase of an ornament being expressed in females that was primarily under selection in malesand would challenge current intra-locus sexual conflict theory (Fisher, 1930; Lande, 1980, 1987;

Bonduriansky and Rowe, 2005), considering that ornament expression is certainly associated with atleast some costs in females.

Although there is growing empirical evidence for male mate choice in polygynous species(for reviews, see Amundsen, 2000; Bonduriansky, 2001; Kraaijeveld et al., 2007), the theoretical basis for thisphenomenon is not trivial because the operational sex ratio is often likely to be male biasedand females are likely to have an equal mating success in such systems. South and colleagues(submitted) explored the possibility that male S. cyaneus may be investing heavily into court-ship and mating and thus be expected to be choosy (Gwynne, 1991, 1993; Johnson and Burley, 1997).They found longevity costs due to courtship and copulation that would decrease a male’sfuture mating success, but whether these costs are large enough to maintain male matechoice in this system is unclear. Servedio and Lande (2006) recently showed that male choosi-ness can be maintained under polygyny even with modest male mating investment. This canoccur when, for example, female signals are not arbitrary but correlate with fecundity orgenetic quality. In S. cyaneus, there is a positive relationship between female paddle areaand body size, which is generally correlated with fecundity in insects (Bonduriansky, 2001). Inaddition, paddle size may signal female age and mating status. Paddle size decreasesmarkedly with age as a result of wear (personal observation) and female lifespan is remarkablylong both in the laboratory (South and Arnqvist, 2008) and in the field (Dégallier et al., 1998).Young, and thus also virgin, females should therefore have on average larger paddles. Anadditional possibility is that males are under selection to direct courtship towards con-specifics to avoid costly heterospecific matings (Servedio, 2007). Sabethes cyaneus is sympatricwith several congeners throughout most of its distribution, and there is striking variationacross species in the number of legs that carry ornaments as well as in their size andcoloration (Lane and Cerqueira, 1942). Thus, males that direct their courtship efforts towardsfemales with large paddles are not only likely to court more fecund females, but alsounmated (i.e. receptive) females of the right species.

South and Arnqvist14

We acknowledge that mutual mate choice should, ideally, be demonstrated by detailedexperimental studies of the role of paddle size during interactions between the sexes. Suchstudies are underway in this system and may provide direct behavioural evidence for mutualmate choice. Nonetheless, by applying geometric morphometrics to explore patterns ofvariation and sexual dimorphism in S. cyaneus, we have gained valuable insights into thelikely selective pressures that act upon the leg paddle ornament in this insect species.Although the data presented here are not entirely conclusive (cf. Bonduriansky and Day, 2003), theypresent a rare and important case in a comparative sense and provide a benchmark forfuture research on systems with sexually monomorphic ornaments.

ACKNOWLEDGEMENTS

Sincerest thanks to R.G. Hancock, W.A. Foster, and R. Taylor for providing us with the S. cyaneuscolony used here and for rearing advice. This study was made possible by financial support from theSwedish Research Council (G.A.) and Zoologiska Stiftelsen (S.S.).

REFERENCES

Alexander, R.D., Marshall, D.C. and Cooley, J.R. 1997. Evolutionary perspectives on insect mating.In The Evolution of Mating Systems in Insects and Arachnids (J.C. Choe and B.J. Crespi, eds.),pp. 4–31. Cambridge: Cambridge University Press.

Allen, B.J. and Levinton, J.S. 2007. Costs of bearing a sexually selected ornamental weapon in afiddler crab. Funct. Ecol., 21: 154–161.

Amundsen, T. 2000. Why are female birds ornamented? Trends Ecol. Evol., 15: 149–155.Andersson, M. 1994. Sexual Selection. Princeton, NJ: Princeton University Press.Armbruster, P. and Hutchinson, R.A. 2002. Pupal mass and wing length as indicators of fecundity in

Aedes albopictus and Aedes geniculatus. J. Med. Entomol., 39: 699–704.Arnqvist, G. and Nilsson, T. 2000. The evolution of polyandry: multiple mating and female fitness in

insects. Anim. Behav., 60: 145–164.Badyaev, A.V. 2004. Developmental perspective on the evolution of sexual ornaments. Evol. Ecol.

Res., 6: 975–991.Beeler, A.E., Rauter, C.M. and Moore, A.J. 2002. Mate discrimination by females in the burying

beetle Nicrophorus orbicollis: the influence of male size on attractiveness to females. Ecol.Entomol., 27: 1–6.

Bonduriansky, R. 2001. The evolution of male mate choice in insects: a synthesis of ideas andevidence. Biol. Rev., 76: 305–339.

Bonduriansky, R. 2007. Sexual selection and allometry: a critical reappraisal of the evidence andideas. Evolution, 61: 838–849.

Bonduriansky, R. and Day, T. 2003. The evolution of static allometry in sexually selected traits.Evolution, 57: 2450–2458.

Bonduriansky, R. and Rowe, L. 2005. Intralocus sexual conflict and the genetic architecture ofsexually dimorphic traits in Prochyliza xanthostoma (Diptera: Piophilidae). Evolution, 59:1965–1975.

Burkhardt, D. and de la Motte, I. 1988. Big ‘antlers’ are favoured: female choice in stalk-eyed flies(Diptera, Insecta), field collected harems and laboratory experiments. J. Comp. Physiol A:Sensory Neural Behav. Physiol., 162: 649–652.

Burley, N. 1981. The evolution of sexual indistinguishability. In Natural Selection and SocialBehaviour: Recent Research and New Theory (R.D. Alexander and D.W. Tinkle, eds.),pp. 121–137. New York: Chiron Press.

Campbell, G. and Tomlinson, A. 1998. The roles of the homeobox genes aristaless and Distal-less inpatterning the legs and wings of Drosophila. Development, 125: 4483–4493.

Allometry of an ornament expressed in both sexes 15

Cook, S.E., Vernon, J.G., Bateson, M. and Guilford, T. 1994. Mate choice in the polymorphicAfrican swallowtail butterfly, Papilio dardanus – male-like females may avoid sexual harassment.Anim. Behav., 47: 389–397.

Cotton, S., Fowler, K. and Pomiankowski, A. 2004. Do sexual ornaments demonstrate heightenedcondition-dependence expression as predicted by the handicap hypothesis? Proc. R. Soc. Lond. B,271: 771–783.

Cotton, S., Rogers, D.W., Small, J., Pomiankowski, A. and Fowler, K. 2006. Variation in preferencefor a male ornament is positively associated with female eyespan in the stalk-eyed fly Diasemopsismeigenii. Proc. R. Soc. Lond. B, 273: 1287–1292.

Coyne, J.A. and Orr, H.A. 2004. Speciation. Sunderland, MA: Sinauer Associates.Cuervo, J.J. and Møller, A.P. 1999. Phenotypic variation and fluctuating asymmetry in sexually

dimorphic feather ornaments in relation to sex and mating system. Biol. J. Linn. Soc., 68: 505-529.Darwin, C. 1871. The Descent of Man, and Sexual Selection in Relation to Sex. London: John Murray.Dégallier, N., Sa, G.C., Monteiro, H.A.O., Castro, F.C., Da Silva, O.V., Brandao, R.C.F.

et al. 1998. Release–recapture experiments with canopy mosquitoes in the genera Haemagogusand Sabethes (Diptera: Culicidae) in Brazilian Amazonia. J. Med. Entomol., 35: 931–936.

Dunn, P.O., Whittingham, L.A. and Pitcher, T.E. 2001. Mating systems, sperm competition, and theevolution of sexual dimorphism in birds. Evolution, 55: 161–175.

Emlen, D.J. 1994. Environmental control of horn length dimorphism in the beetle Onthophagusacuminatus (Coleoptera: Scarabaeidae). Proc. R. Soc. Lond. B, 256: 131–136.

Emlen, D.J. and Nijhout, H.F. 2000. The development and evolution of exaggerated morphologies ininsects. Annu. Rev. Entomol., 45: 661–708.

Emlen, D.J., Lavine, L.C. and Ewen-Campen, B. 2007. On the origin and evolutionary diversificationof beetle horns. Proc. Natl. Acad. Sci. USA, 104: 8661–8668.

Evans, M.R. 2004. Limits on the evolution of tail ornamentation in birds. Am. Nat., 163: 341–357.Ferson, S., Rohlf, F.J. and Koehn, R.K. 1985. Measuring shape variation of two-dimensional

outlines. Syst. Zool., 34: 59–68.Fisher, R.A. 1930. The Genetical Theory of Natural Selection. Oxford: Clarendon Press.Forbes, M.R., Schalk, G., Miller, J.G. and Richardson, J.M.L. 1997. Male–female morph

interactions in the damselfly Nehalennia irene (Hagen). Can. J. Zool., 75: 253–260.Godin, J.G.J. and McDonough, H.E. 2003. Predator preference for brightly colored males in the

guppy: a viability cost for a sexually selected trait. Behav. Ecol., 14: 194–200.Grafen, A. 1990a. Biological signals as handicaps. J. Theor. Biol., 144: 517–546.Grafen, A. 1990b. Sexual selection unhandicapped by the Fisher process. J. Theor. Biol., 144:

475–516.Green, A.J. 1992. Positive allometry is likely with mate choice, competitive display and other

functions. Anim. Behav., 43: 170–172.Gwynne, D.T. 1991. Sexual competition among females: what causes courtship-role reversal? Trends

Ecol. Evol., 6: 118–121.Gwynne, D.T. 1993. Food quality controls sexual selection in mormon crickets by altering male

mating investment. Ecology, 74: 1406–1413.Hancock, R.G., Foster, W.A. and Yee, W.L. 1990a. Courtship behavior of the mosquito Sabethes

cyaneus (Diptera: Culicidae). J. Insect Behav., 3: 401–416.Hancock, R.G., Yee, W.L. and Foster, W.A. 1990b. Tests of Sabethes cyaneus leg paddle function in

mating and flight. J. Am. Mosquito Control Assoc., 6: 733–735.Harbach, R.E. and Petersen, J.L. 1992. Two species previously confused under the concept of

Sabethes tarsopus in Central America (Diptera: Culicidae). Mosquito Syst., 24: 102–124.Hingle, A., Fowler, K. and Pomiankowski, A. 2001. Size-dependent female mate preference in the

stalk-eyed fly, Cyrtodiopsis dalmanni. Anim. Behav., 61: 589–595.Hooper, P.L. and Miller, G.F. 2008. Mutual mate choice can drive costly signaling even under perfect

monogamy. Adaptive Behav., 16: 53–69.

South and Arnqvist16

Huxley, J.S. 1914. The courtship habits of the great crested grebe (Podiceps cristatus); with andaddition to the theory of sexual selection. Proc. Zool. Soc. Lond., 35: 491–562.

Huxley, J.S. 1932. Problems of Relative Growth. New York: Lincoln Mac Veagh – The Dial Press.Jang, Y. and Greenfield, M.D. 1996. Ultrasonic communication and sexual selection in wax

moths: female choice based on energy and asynchrony of male signals. Anim. Behav., 51:1095–1106.

Jang, Y. and Greenfield, M.D. 1998. Absolute versus relative measurements of sexual selection:assessing the contributions of ultrasonic signal characters to mate attraction in lesser wax moths,Achroia grisella (Lepidoptera: Pyralidae). Evolution, 52: 1383–1393.

Jennions, M.D. 1993. Female choice in birds and the cost of long tails. Trends Ecol. Evol., 8:230–232.

Johnson, K. and Burley, N.T. 1997. Mating tactics and mating systems of birds. Ornithol. Monogr.,1997: 21–60.

Jones, I.L. and Hunter, F.M. 1993. Mutual sexual selection in a monogamous seabird. Nature, 362:238–239.

Jones, I.L. and Hunter, F.M. 1999. Experimental evidence for mutual inter- and intrasexual selectionfavouring a crested auklet ornament. Anim. Behav., 57: 521–528.

Judd, D.D. 1996. Review of the systematics and phylogenetic relationships of Sabethini (Diptera:Culicidae). Syst. Entomol., 21: 129–150.

Kemp, D.J. 2006. Heightened phenotypic variation and age-based fading of ultraviolet butterflywing coloration. Evol. Ecol. Res., 8: 515–527.

Kemp, D.J. and Rutowski, R.L. 2007. Condition dependence, quantitative genetics, and the potentialsignal content of iridescent ultraviolet butterfly colorations. Evolution, 61: 168–183.

Kimball, R.T. and Ligon, J.D. 1999. Evolution of avian plumage dichromatism from a proximateperspective. Am. Nat., 154: 182–193.

Klingenberg, C.P. 1996. Multivariate allometry. In Advances in Morphometrics (L.F. Marcus,M. Corti, A. Loy, G.J.P. Naylor, and D.E. Slice, eds.), pp. 23–49. New York: Plenum Press.

Klingenberg, C.P. and Zimmermann, M. 1992. Static, ontogenetic, and evolutionary allometry: amultivariate comparison in nine species of water striders. Am. Nat., 140: 601–620.

Knell, R.J., Pomfret, J.C. and Tomkins, J.L. 2004. The limits of elaboration: curved allometriesreveal the constraints on mandible size in stag beetles. Proc. R. Soc. Lond. B, 271: 523–528.

Kodric-Brown, A., Sibly, R.M. and Brown, J.H. 2006. The allometry of ornaments and weapons.Proc. Natl. Acad. Sci. USA, 103: 8733–8738.

Kraaijeveld, K., Gregurke, J., Hall, C., Komdeur, J. and Mulder, R.A. 2004. Mutual ornamentation,sexual selection, and social dominance in the black swan. Behav. Ecol., 15: 380–389.

Kraaijeveld, K., Kraaijeveld, F.J.L. and Komdeur, J. 2007. The evolution of mutual ornamentation.Anim. Behav., 74: 657–677.

Land, M.F., Gibson, G., Horwood, J. and Zeil, J. 1999. Fundamental differences in the opticalstructure of the eyes of nocturnal and diurnal mosquitoes. J. Comp. Physiol A: Sensory NeuralBehav. Physiol., 185: 91–103.

Lande, R. 1980. Sexual dimorphism, sexual selection, and adaptation in polygenic characters.Evolution, 34: 292–305.

Lande, R. 1987. Genetic correlations between the sexes in the evolution of sexual dimorphism andmating preferences. In Sexual Selection: Testing the Alternatives (J.W. Bradbury and M.B.Andersson, eds.), pp. 83–95. Chichester: Wiley.

Lande, R. and Arnold, S.J. 1985. Evolution of mating preference and sexual dimorphism. J. Theor.Biol., 117: 651–664.

Lane, J. and Cerqueira, N.L. 1942. Os sabetineos da América (Diptera: Culicidae). Arquivos deZoologia, 3: 473–849.

Langmore, N.E. and Bennett, A.T.D. 1999. Strategic concealment of sexual identity in an estrildidfinch. Proc. R. Soc. Lond. B, 266: 543–550.

Allometry of an ornament expressed in both sexes 17

Ligon, J.D. 1999. Evolution of Avian Breeding Systems. Oxford: Oxford University Press.McKinney, M.L. and McNamara, K.J. 1991. Heterochrony: Evolution of Ontogeny. New York:

Plenum Press.Munoz, A., Aparicio, J.M. and Bonal, R. 2008. Male barn swallows use differential resource

allocation rules to produce ornamental tail feathers. Behav. Ecol., 19: 404–409.Okazawa, T., Horio, M., Suzuki, H. and Mogi, M. 1986. Colonization and laboratory bionomics of

Topomyia yanbarensis (Diptera: Culicidae). J. Med. Entomol., 23: 493–501.Petrie, M. 1992. Are all secondary sexual display structures positively allometric and, if so, why?

Anim. Behav., 43: 173–175.Philips, T.K., Hancock, R.G. and Foster, W.A. 1996. Epigamic display and unique mating position in

Wyeomyia arthrostigma (Diptera: Culicidae). J. Insect Behav., 9: 739–753.Polak, M. and Starmer, W.T. 2005. Environmental origins of sexually selected variation and a

critique of the fluctuating asymmetry-sexual selection hypothesis. Evolution, 59: 577–585.Rauskolb, C. 2001. The establishment of segmentation in the Drosophila leg. Development, 128:

4511–4521.Ricker, W.E. 1984. Computation and uses of general trend lines. Can. J. Zool., 62: 1897–1905.Rohlf, F.J. 1992. The analysis of shape variation using ordinations of fitted functions. In Ordination

in the Study of Morphology, Evolution and Systematics of Insects (J.T. Sorensen and R. Foottit,eds.), pp. 95–112. Amsterdam: Elsevier.

Rohlf, F.J. and Archie, J.A. 1984. Comparison of Fourier methods for the description of wing shapein mosquitoes (Diptera: Culicidae). Syst. Zool., 33: 302–317.

Rowe, L. and Houle, D. 1996. The lek paradox and the capture of genetic variance by conditiondependent traits. Proc. R. Soc. Lond. B, 263: 1415–1421.

Sawada, M. 1999. ModelII: Geometric Mean Regression Add-In for Excel ’97. Freeware byM. Sawada, University of Ottawa.

Schluter, D. and Price, T. 1993. Honesty, perception and population divergence in sexually selectedtraits. Proc. R. Soc. Lond. B, 253: 117–122.

Servedio, M.R. 2007. Male versus female mate choice: sexual selection and the evolution of speciesrecognition via reinforcement. Evolution, 61: 2772–2789.

Servedio, M.R. and Lande, R. 2006. Population genetic models of male and mutual mate choice.Evolution, 60: 674–685.

Shannon, R.C. 1931. On the classification of Brazilian Culicidae with special reference to thosecapable of harboring the yellow fever virus. Proc. Entomol. Soc. Wash., 33: 125–164.

Shuster, S.M. and Wade, M.J. 2003. Mating Systems and Strategies. Princeton, NJ: PrincetonUniversity Press.

Siegel, J.P., Novak, R.J., Lampman, R.L. and Steinly, B.A. 1992. Statistical appraisal of theweight–wing length relationship of mosquitos. J. Med. Entomol., 29: 711–714.

Simmons, L.W. 1986. Female choice in the field cricket Gryllus bimaculatus (De Geer). Anim. Behav.,34: 1463–1470.

Simmons, L.W. and Tomkins, J.L. 1996. Sexual selection and the allometry of earwig forceps. Evol.Ecol., 10: 97–104.

Sivinski, J. 1997. Ornaments in the Diptera. Florida Entomol., 80: 142–164.Slice, D.E. 2002. Morpheus et al. – Multiplatform Software for Morphometrics Analysis.Sokal, R.R. and Rohlf, F.J. 1995. Biometry: The Principles and Practice of Statistics in Biological

Research. New York: W.H. Freeman.South, S.H. and Arnqvist, G. 2008. Evidence of monandry in a mosquito (Sabethes cyaneus) with

elaborate ornaments in both sexes. J. Insect Behav., 21: 451–459.Storey, J.D. 2003. The positive false discovery rate: A Bayesian interpretation and the q-value. Ann.

Stat., 31: 2013–2035.Swaddle, J.P. 2003. Fluctuating asymmetry, animal behavior, and evolution. Adv. Study Behav., 32:

169–205.

South and Arnqvist18

Thornhill, R. and Alcock, J. 1983. The Evolution of Insect Mating Systems. Cambridge, MA:Harvard University Press.

Trivers, R.L. 1972. Parental investment and sexual selection. In Sexual Selection and the Descent ofMan (B. Campbell, ed.), pp. 136–179. London: Heinemann.

Van Gossum, H., Beirinckx, K., Forbes, M.R. and Sherratt, T.N. 2007. Do current hypothesesexplain continental and seasonal variation in female morph frequencies of the damselfly,Nehalennia irene? Biol. J. Linn. Soc., 90: 501–508.

Watson, P.J. and Thornhill, R. 1994. Fluctuating asymmetry and sexual selection. Trends Ecol. Evol.,9: 21–24.

West-Eberhard, M.J. 1983. Sexual selection, social competition, and speciation. Q. Rev. Biol., 58:155–183.

Wiley, R.H. and Poston, J. 1996. Indirect mate choice, competition for mates, and coevolution of thesexes. Evolution, 50: 1371–1381.

Wilkinson, G.S. and Dodson, G.N. 1997. Function and evolution of antlers and eye-stalks in flies.In Mating Systems in Insects and Arachnids (J.C. Choe and B.J. Crespi, eds.), pp. 310–328.Cambridge: Cambridge University Press.

Wilkinson, G.S. and Reillo, P.R. 1994. Female choice response to artificial selection on anexaggerated male trait in a stalk-eyed fly. Proc. R. Soc. Lond. B, 255: 1–6.

Zsemlye, J.L., Hancock, R.G. and Foster, W.A. 2005. Analysis of a complex vertical copulatory-courtship display in the yellow fever vector Sabethes chloropterus. Med. Vet. Entomol., 19:276–285.

APPENDICES

Appendix 1. The percentage of variance explained by each of the 10 firstprincipal components of variation in leg paddle shape

Principal component Percentage of variance explained

Tibial paddle 1 57.6102 13.0153 9.1454 4.8065 4.2106 2.8097 2.4818 1.6199 1.047

10 0.616

Tarsal paddle 1 52.2782 16.1073 9.5844 5.4215 4.1396 2.2557 1.9148 1.6279 1.290

10 0.884

Allometry of an ornament expressed in both sexes 19

Appendix 2. Repeatabilities, estimated as Pearson’s correlation coefficients, of all measurements taken

Trait Side Female Male Joint

Wing Left 0.87 0.70 0.87Thorax Left 0.62 0.73 0.66Proboscis Left 0.87 0.60 0.78Antenna Left 0.66 0.64 0.87Fore femur Left 0.89 0.81 0.87Fore tibia Left 0.93 0.77 0.87Fore tarsus Left 0.96 0.87 0.95Hind femur Left 0.94 0.70 0.90Hind tibia Left 0.91 0.73 0.88Hind tarsus Left 0.99 0.74 0.94Mid femur Left 0.96 0.76 0.92Mid tibia Left 0.81 0.67 0.82Mid tarsus Left 0.73 0.64 0.93

Tibial paddleArea Left 0.98 0.97 0.98Shape PC1 Left 0.98 0.97 0.98Shape PC2 Left 0.76 0.92 0.85Shape PC3 Left 0.78 0.91 0.88Shape PC4 Left 0.80 0.66 0.78Shape PC5 Left 0.51 0.70 0.59Area Right 0.98 0.90 0.97Shape PC1 Right 0.98 0.99 0.98Shape PC2 Right 0.92 0.83 0.88Shape PC3 Right 0.80 0.96 0.93Shape PC4 Right 0.71 0.96 0.91Shape PC5 Right 0.72 0.88 0.80Asymmetry Area N/A 0.97 0.91 0.94Asymmetry Shape PC1 N/A 0.97 0.97 0.97Asymmetry Shape PC2 N/A 0.72 0.77 0.74Asymmetry Shape PC3 N/A 0.82 0.94 0.89Asymmetry Shape PC4 N/A 0.80 0.93 0.89Asymmetry Shape PC5 N/A 0.67 0.75 0.72

Tarsal paddleArea Left 0.94 0.92 0.96Shape PC1 Left 0.90 0.95 0.92Shape PC2 Left 0.70 0.88 0.84Shape PC3 Left 0.68 0.34 0.47Shape PC4 Left 0.58 0.59 0.62Shape PC5 Left 0.72 0.81 0.76Area Right 0.97 0.90 0.97Shape PC1 Right 0.96 0.96 0.96Shape PC2 Right 0.86 0.86 0.88Shape PC3 Right 0.78 0.76 0.78Shape PC4 Right 0.63 0.67 0.66Shape PC5 Right 0.78 0.87 0.84Asymmetry Area N/A 0.79 0.76 0.78Asymmetry Shape PC1 N/A 0.87 0.88 0.87Asymmetry Shape PC2 N/A 0.84 0.87 0.85Asymmetry Shape PC3 N/A 0.59 0.48 0.51Asymmetry Shape PC4 N/A 0.39 0.58 0.54Asymmetry Shape PC5 N/A 0.68 0.86 0.78

South and Arnqvist20

Appendix 3. Tests of directional asymmetry in components of leg paddle morphology

Trait Sex Mean t d.f. P

Tibial paddleArea Joint 0.002 0.267 159 0.790Shape PC1 Joint −0.002 −0.256 159 0.800Shape PC2 Joint −0.001 −0.354 159 0.724Shape PC3 Joint −0.004 −1.162 159 0.247Shape PC4 Joint 0.003 1.325 159 0.187Shape PC5 Joint 0.001 0.556 159 0.579

Tarsal paddleArea Joint 0.004 0.859 159 0.391Shape PC1 Joint 0.008 1.230 159 0.220Shape PC2 Joint −0.001 −0.178 159 0.859Shape PC3 Joint −0.001 −0.291 159 0.771Shape PC4 Joint 0.005 2.119 159 0.036Shape PC5 Joint −0.004 −1.765 159 0.080

Tibial paddleArea Female −0.006 −0.463 79 0.645Shape PC1 Female 0.015 1.195 79 0.236Shape PC2 Female 0.001 0.116 79 0.908Shape PC3 Female −0.007 −1.670 79 0.099Shape PC4 Female 0.003 0.929 79 0.356Shape PC5 Female 0.003 0.838 79 0.405

Tarsal paddleArea Female −0.004 −0.568 79 0.571Shape PC1 Female 0.001 0.085 79 0.933Shape PC2 Female 0.007 1.318 79 0.191Shape PC3 Female −0.006 −1.235 79 0.221Shape PC4 Female 0.006 1.786 79 0.078Shape PC5 Female −0.003 −1.165 79 0.247

Tibial paddleArea Male 0.011 0.942 79 0.349Shape PC1 Male −0.020 −1.699 79 0.093Shape PC2 Male −0.003 −0.715 79 0.477Shape PC3 Male −0.000 −0.070 79 0.945Shape PC4 Male 0.004 0.967 79 0.337Shape PC5 Male −0.000 −0.044 79 0.965

Tarsal paddleArea Male 0.011 2.193 79 0.031Shape PC1 Male 0.015 1.584 79 0.117Shape PC2 Male −0.008 −1.157 79 0.250Shape PC3 Male 0.004 0.741 79 0.461Shape PC4 Male 0.004 1.243 79 0.218Shape PC5 Male −0.005 −1.319 79 0.191

Note: Significant values (P = 0.05; no compensation for multiple tests) are highlighted in bold.

Allometry of an ornament expressed in both sexes 21